The linked paper is a little underwhelming: they find that people generally find a similarity-based tweet recommendation engine "interesting" even with the secret twist that it only shows tweets from those who have opposing views on abortion. Given the study wasn't focused on politics or controversy (except to the extent Chilean Twitter feeds in general are) and no attempt was made to gauge whether anyone had felt their views challenged or found interesting new perspectives, I don't think the "filter bubble" was really burst at all.